The lactol route to fesoterodine: An amine-promoted Friedel-Crafts alkylation on commercial scale

Article abstract of DOI:10.1021/op200107g

We report the discovery and optimization of an amine-promoted Friedel-Crafts alkylation of cinnamaldehyde with 4-hydroxymethyl phenol. This reaction has been used successfully on commercial scale (200 kg) in the context of the manufacture of fesoterodine, a muscarinic antagonist used for the treatment of overactive bladder. Reductive aminations of diisopropylamine and lactol 4 are also discussed, as well as the resolution of the racemic amine rac-2 into its enantiomerically pure form.

Full text of DOI:10.1021/op200107g

ARTICLE
pubs.acs.org/OPRD
The Lactol Route to Fesoterodine: An Amine-Promoted FriedelÀCrafts
Alkylation on Commercial Scale
Olivier Dirat,* Andrew J. Bibb, Colin M. Burns, Graham D. Checksfield, Barry R. Dillon, Stuart E. Field,
Steven J. Fussell, Stuart P. Green, Clive Mason, Jinu Mathew, Suju Mathew, Ian B. Moses,
Petar I. Nikiforov, Alan J. Pettman, and Flavien Susanne
Department of Chemical Research and Development, Pfizer Global Research and Development, Ramsgate Road, Sandwich, Kent,
CT13 9NJ, U.K.
ABSTRACT: We report the discovery and optimization of an amine-promoted FriedelÀCrafts alkylation of cinnamaldehyde with
4-hydroxymethyl phenol. This reaction has been used successfully on commercial scale (200 kg) in the context of the manufacture of
fesoterodine, a muscarinic antagonist used for the treatment of overactive bladder. Reductive aminations of diisopropylamine and
lactol 4 are also discussed, as well as the resolution of the racemic amine rac-2 into its enantiomerically pure form.
’ INTRODUCTION
Fesoterodine 1 is a muscarinic antagonist used for the treat-
ment of overactive bladder1 and is commercialized under the
name Toviaz. Fesoterodine is a pro-drug of 5-hydroxymethyl
tolterodine 2, which is the active metabolite of tolterodine 3, a
muscarinic antagonist commercialized under the name Detrol
(Figure 1). Despite the close structural resemblance, fesotero-
dine has been shown to display superior efficacy and tolerability
over tolterodine.2
The existing process for the commercial manufacture of
Figure 1. Structure of fesoterodine and tolterodine.
fesoterodine was linear (11 steps) and utilized difficult to handle
reagents (e.g., LiAlH4) on commercial production scale.3 For
FriedelÀCrafts alkylation of cinnamaldehyde 6 with the unpro-
these reasons, a more efficient and plant-friendly process was
sought.
tected phenol 5 (Figure 2).
Amine-Catalyzed FriedelÀCrafts Alkylation. Due to the
facile dehydration of 5, Lewis or Brønsted acids could not be
used to activate 6 towards FriedelÀCrafts alkylation with 5, so we
All previous synthetic approaches to fesoterodine involved
We were encouraged by recent reports of iminium activation of
’ RESULTS AND DISCUSSION
turned our attention to alternative ways to facilitate the reaction.
multiple chemical transformations in order to install the hydro-
xymethyl functionality (use of protecting groups, redox chem-
istry, or functional group manipulations). These long synthetic
routes were designed because the hydroxymethyl group is
R,β-unsaturated ketones and aldehydes towards enantioselective
FriedelÀCrafts alkylation reactions with electon-rich arenes.5
Accordingly, we tested up to stoichiometric amounts of com-
mercially available proline- and imidazolidinone- based chiral
organocatalysts but failed to observe any desired product even
under refluxing THF conditions. We reasoned that this was
unusually prone to dehydration to form quinone methide 20
due to a synergistic effect of the phenol substituent in the para-
position.4 The resulting activated electrophile readily reacts with
probably due to the lack of reactivity of 5, and the steric
a wide range of nucleophiles to form impurities. As we aimed to
hindrance of the chiral amines used. Achiral aminocatalysis was
achieve a more concise synthesis of fesoterodine, we focused our
pioneered by Knoevenagel in 1896 and has since been mostly
used for aldol reactions.6 Although we were unaware of examples
efforts on approaches avoiding the use of protecting groups, red-
ox chemistry, or functional group manipulations. In order to
achieve this goal, the hydroxymethyl group needs to be present
from the start of the synthesis in the unprotected form, thus
of amine-catalyzed FriedelÀCrafts alkylations, we wondered
whether this approach (Scheme 1) could be useful for the
synthesis of fesoterodine.
limiting the types of reaction conditions compatible with this
We were delighted to observe good to excellent yields with
reactive group (e.g., avoid the use of Brønsted or Lewis acids).
Accordingly, we envisioned that fesoterodine could be prepared
from the advanced intermediate 2, common with the existing
process. Intermediate 2 would be prepared from lactol 4 via
stoichiometric amounts of a range of cyclic secondary amines
when cinnamaldehyde 6 and phenol 5 were heated to reflux in
reductive amination. Finally, we envisioned that the most direct
and atom economical way to synthesize 4 would be via a
Received: April 19, 2011
Published: June 17, 2011
r
2011 American Chemical Society
1010
dx.doi.org/10.1021/op200107g Org. Process Res. Dev. 2011, 15, 1010–1017
|

Organic Process Research & Development
ARTICLE
converted to the desired aminal 8 over the course of 3 h
(Scheme 2). The formation of this 2:1 adduct may prevent
cinnamaldehyde from degrading under the reaction conditions
hence explaining the optimal results obtained with this stoichio-
metry. The formation of this adduct also generates water, which
can react with the iminium species 10 to provide β-hydroxyalde-
hyde 11. This side product is not stable under the reaction
conditions and undergoes a retro-aldol reaction to produce
benzaldehyde 12 (observed) and presumably acetaldehyde.
About 20 mol % of benzaldehyde 12 was typically generated
during the reaction, but this undesired pathway can be almost
completely suppressed by the use of DeanÀStark conditions to
remove water as it is generated.
Figure 2. Retrosynthetic analysis of fesoterodine.
Scheme 1. Amino-Catalyzed FriedelÀCrafts Alkylation
The condensation of cinnamaldehyde and 4-hydroxymethyl
phenol is not completely regioselective, and about 10% of regio-
isomer 13 was formed. Aminal 8 was not isolated but was readily
hydrolysed using 2 M HCl, and lactol 4 was isolated as a cream
solid by crystallization from a mixture of ethyl acetate and
toluene.7 This crystallization was pivotal to obtaining the re-
quired quality of 4 as the ingoing process stream contained large
amounts of excess reagents, impurities, and polymeric materials.
The solvent composition (toluene/ethyl acetate 4:1) of this
crystallization results from balancing product recovery and
impurity purge with toluene acting as antisolvent and ethyl
acetate as solvent. These conditions resulted in an excellent
purge of reagents, impurities, and polymeric materials and
produced very pure lactol 4 with about 20% product lost to the
liquor. This process has been repeated successfully on 200 kg
scale in 57% yield on many occasions.
Table 1. Amine Screen for FridelÀCrafts Alkylation in
Refluxing tert-Amyl Alcohol
Entry
Amine
Yielda
1
Diisopropylamine
Dibenzylamine
Di-n-butylamine
Piperidine
0%b
10%c
34%c
73%
69%
55%
80%
36%d
53%
20%c
2
3
4
Reductive Amination. Reductive amination was now re-
quired in order to install the diisopropylamine moiety of
fesoterodine. Four products were typically observed during the
development of this transformation: the desired product rac-2,
racemic tolterodine rac-3, triol 14, and unreacted lactol 4
(Table 2). Following the reductive amination step, the reaction
mixture is combined with a chiral acid to effect the resolution of
the enantiomers of 2. Consequently, the molecules that do not
contain basic nitrogen (triol 14 and lactol 4) would be expected
to be purged to the mother liquor during the isolation. The
challenges of the reductive amination are several fold: (i)
conversion; (ii) selectivity between direct reduction of lactol 4
into triol 14 and desired reductive amination; and (iii) the
control of over-reduction of rac-2 into rac-3 as this impurity is
poorly purged downstream due to the close structural similarity.
Results from Table 2 show that heterogeneous hydrogenation
using Pd/C as catalyst was the most selective way found to
reductively aminate diisopropylamine with lactol 4 (entry 1). Pt/C
gave poor selectivity for over-reduction to racemic tolterodine
rac-3 and direct reduction of lactol to give triol 14, even at partial
conversion (entry 2). Ru/C was selective towards undesired
direct reduction of lactol 4 (entry 3). Homogenous transfer
hydrogenation provided selectively 14 with no pretreatment with
Ti(OiPr)4, and significant amounts of over-reduction with pre-
treatment even at low conversion (entries 4À5). Pretreatment
with MgSO4 was ineffective at facilitating the condensation of
diisopropylamine with 4 (entry 6), and NaBH4 produced large
amounts of over-reduction (entry 7). Finally, NaBH3(CN) with
pretreatment with Ti(OiPr)4 selectively gave the over-reduced
product at 70 °C but produced mostly the desired product at
room temperature (entries 8À9).
5
Piperazine
6
Morpholine
7
N-methyl piperazine
N-acetyl piperazine
Thiopiperazine
N-isopropyl piperazine
8
9
10
a Yield determined by quantitative HPLC after hydrolysis. b No conver-
sion observed. c Low conversion. d Multiple products observed.
tert-amyl alcohol (Table 1). Substoichiometric amounts of amine
led to stalling of the reaction in direct proportion with the
amount of amine used. The reaction did not proceed to any
significant extent at temperatures below 80 °C, and the preferred
solvents were found to be toluene, tert-amyl alcohol, and
chlorobenzene. Results from Table 1 show that this reaction is
quite sensitive to both electronic and steric factors on the amine.
Acyclic amines (entries 1À3) provided no or low conversions.
We were particularly interested in diisopropylamine, since if this
transformation had worked, reduction of the resultant aminal
would have yielded directly rac-2 without the need to perform a
separate reductive amination. Unfortunately this amine is too
hindered to be an effective promoter and no conversion was
observed. Cyclic amines (entries 4À10) showed the best con-
versions, with N-methyl piperazine (entry 7) and piperidine
(entry 4) being the most efficient. Due to the presence of the
unprotected phenol, the amine is incorporated in the aminal
intermediate 8 hence requiring the use of at least 1 equiv of
amine. Further optimization revealed that a 2:1 ratio of amine to
cinnamaldehyde provided higher yields. The probable explana-
tion of this fact was found when the reaction was performed
1
in deuterated toluene and followed by H NMR as it revealed
Reductive amination using heterogeneous hydrogenation with
Pd/C was optimized further. Methanol was the only solvent that
the immediate formation of a 2:1 adduct 9, which then slowly
1011
dx.doi.org/10.1021/op200107g |Org. Process Res. Dev. 2011, 15, 1010–1017

Organic Process Research & Development
ARTICLE
Scheme 2. Proposed Mechanism
Table 2. Reductive Amination of Diisopropylamine and 4
(Figure 3A and B). The initial rate of reaction is influenced by the
concentration of both of these starting materials, hence suggest-
ing that they are both involved in the rate-limiting step of this
reaction. The initial rate was then studied as a function of the
catalyst loading (Figure 3C). The results suggest that catalyst is
involved also in the rate-limiting step and that, beyond 10% w/w
of catalyst charge, extra catalyst would not significantly increase
the overall rate of reaction. Finally, we studied the initial rate of
reaction as a function of hydrogen pressure (Scheme 3D) and
found that hydrogen pressure did not have a significant effect on
the reaction rates. This result could be the consequence of two
possible causes: (i) the overall rate of reaction is limited by the
formation of the hydrogenation substrate and not by the hydro-
genation event, or (ii) that the hydrogenation event is the rate-
limiting step but does not depend on hydrogen pressure, which
can be the result of a LangmuirÀHinshelwood type kinetics.8
These kinetic studies therefore suggest that the overall reaction
rate is a function of the concentrations of lactol 4, diisopropy-
lamine, and the Pd catalyst. Since the reaction time is long (20 h),
it is unlikely that the hydrogenation event is limited by the
efficiency of mass transfer within the three-phase system and
therefore changes in the parameters affecting mass transfer (e.g.,
mixing) during scale-up should not impact the overall rate of
reaction, so long as mass transfer is high enough to allow the
reaction to proceed under kinetic control. This was proven to be
the case across a wide range of scales, from 1 g to 200 kg.
In order to probe the structure of the hydrogenation substrate-
(s), the reaction was performed using deuterium instead of
hydrogen, and the result (mass spectrometry) showed that only
one deuterium atom was incorporated in the product. This
suggests that the hydrogenation substrate was iminium ion 15
or aminal 17 or 24, as opposed to enamine 16 (Scheme 3). NMR
studies of the reaction mixture in deuterated methanol prior to
adding the Pd/C catalyst and applying hydrogen showed that
Entry
1
Conditions
Pd/C, MeOH,
rac-2
rac-3
14
4
93
1
1
5
115 psi, 40 °C
2
3
Pt/C, t-Amyl-OH, 100 psi, 80 °C
Ru/C, t-Amyl-OH,
100 psi, 80 °C
50
5
5
1
10
96
35
0
4
[RuCl2p-Cymene]2,
Diamine, THF,
HCOOH, rt
1
0
8
99
0
0
5a
[RuCl2p-Cymene]2,
Diamine, THF,
HCOOH, rt
46
46
6b
7a
8a
9a
NaBH4, rt
0
70
1
0
30
99
4
100
0
0
0
0
8
NaBH4, rt
NaBH3CN, 70 °C
NaBH3CN, rt
0
88
0
a Pretreatment with Ti(OiPr)4, THF, reflux, 1 h. b Pretreatment with
Mg(SO)4, THF, rt, 1 h.
provided good conversion and selectivity over formation of rac-3.
In order to gain insights into the mechanism of the reaction
and to assess the scale-up risks, some kinetic studies were per-
formed (Figure 3). First we studied the initial rate of reaction as a
function of the concentration of lactol 4 and diisopropylamine
1012
dx.doi.org/10.1021/op200107g |Org. Process Res. Dev. 2011, 15, 1010–1017

Organic Process Research & Development
ARTICLE
Figure 3. Kinetic studies: plots of initial rate of reaction vs. reaction parameters. (A) Initial rate (mmol/min) as a function of lactol 4 concentration
(mol/L). (B) Initial rate (mmol/min) as a function of diisopropylamine concentration (mol/L). (C) Initial rate (mmol/min) as a function of catalyst
loading (w/w %). (D) Initial rate (mmol/min) as a function of hydrogen pressure (psi).
Scheme 3. Possible Hydrogenation Substrates and Proposed Mechanism
protons in the beta position from the nitrogen (around 2 ppm)
in lactol 4 were exchanged by deuterium, thus indicating that
enamine 16 was indeed formed in the reaction mixture (Figure 4).
These studies however failed to observe 15, 16, 17, or 24,
presumably due to their low concentration. Finally, we searched
for a set of conditions that would provide the widest window for
reaching high conversion with low amounts of over-reduction.
DoE experimentation showed that the best compromise was
1013
dx.doi.org/10.1021/op200107g |Org. Process Res. Dev. 2011, 15, 1010–1017

Organic Process Research & Development
ARTICLE
Figure 4. 1H NMR in MeOH-d4 of Lactol 4 with and without diisopropylamine.
Scheme 4. Resolution of rac-2
Scheme 5. Completion of the Synthesis of Fesoterodinea
found using 10% w/w of catalyst at low temperatures (40 °C)
and high stoichiometry of diisopropylamine (3 equiv), presum-
ably in order to shift the pre-equilibrium towards 15, 17, or 24.
After filtration to remove the catalyst, the hydrogenation reaction
stream was not isolated but telescoped directly into the next step
(resolution) after a solvent switch into tert-amyl alcohol via
distillation under reduced pressure.
Resolution of the Enantiomers of 2. In order to resolve the
enantiomers of 2, chiral acids were screened in order to identify
those that would allow a selective crystallization of the desired
diastereomeric salt. (R)-Acetyl mandelic acid was identified as a
promising lead and provided on optimization an efficient crystal-
lization of the desired salt 18 when no more than 0.5 equiv of
chiral acid in tert-amyl alcohol was used (Scheme 4). Methanol
and diisopropylamine present in the ingoing solution of rac-2
from the previous step had to be removed to low levels since the
presence of methanol leads to the formation of the methyl ether
impurity 19,9 and diisopropylamine selectively forms a salt with
acetyl mandelic acid, thereby reducing the available acid stoichi-
ometry for the desired salt formation. Accordingly, a distillation
under reduced pressure was performed at the end of the
hydrogenation reaction. Under those conditions, the resolution
step afforded 38% yield (80% of theoretical yield) of a salt
containing 2 with >99% ee and complete purge of lactol 4 and
triol 14.
a Reagents and conditions: (i) K2CO3, water, toluene, 50 °C, 80%; (ii)
isobutyryl chloride, DCM, À10 °C, 90%; (iii) fumaric acid, MEK, 82%.
common intermediate 2 after a seeded crystallization from
toluene in 85% yield. Acylation of the phenol using isobutyryl
chloride was accomplished in dichloromethane at À10 °C
(Scheme 5). The selectivity of the acylation (mono (1 and 21)
vs diacylation (22) and acylation of phenol (1) vs primary
alcohol (21)) is influenced by solvent, temperature, and stoichi-
ometry of isobutyryl chloride (Figure 5). The internal nitrogen
acts as a base in this reaction, and the addition of an extra
equivalent of base leads to poorer selectivities. The free base of 1
is not isolated, and the organic stream is telescoped into the next
step. The synthesis is completed by a fumarate salt formation in
methyl ethyl ketone to afford the commercial salt of fesoterodine
in 82% yield.
Conclusion. This process illustrates the discovery and use on
commercial scale (200 kg batch size) of an amine-promoted
FriedelÀCrafts alkylation and offers a significantly shorter
Completion of the Synthesis. The synthesis of fesoterodine
was completed by a salt break of 18 in a biphasic mixture of
toluene and aqueous potassium carbonate, yielding the advanced
1014
dx.doi.org/10.1021/op200107g |Org. Process Res. Dev. 2011, 15, 1010–1017

Organic Process Research & Development
ARTICLE
stirred in methanol (1.5 L, 7.5 mL/g). Diisopropylamine (237 g,
2.34 mol, 3.0 equiv) was then added over 15 min maintaining the
temperature below 40 °C. The resulting solution was then stirred
for 1 h under nitrogen. The catalyst Pd-ESCAT 142 (Supplier
Engelhard) [(5% w/w Pd/C paste, ca. 50% water wet) 20 g, 10%
w/w] was added, and the system was purged with nitrogen. The
mixture was hydrogenated at 793 kPa (115 psi, 7.92 bar) at a
temperature of 40 °C for 20 h. The mixture was cooled and
purged with nitrogen and then filtered using filter aid, and the
residue pad was washed with methanol (2 Â 400 mL, 2 Â 2 mL/g).
The combined filtrate and washings were transferred to a
distillation vessel where the product solution was concentrated to
a 330 mL (1.65 mL/g) volume under reduced pressure. tert-Amyl
alcohol (670 mL, 3.35 mL/g) was added, and the mixture was
reconcentrated to 330 mL (1.65 mL/g) volume. Six further
additions of tert-amyl alcohol (each of 670 mL, 3.35 mL/g), each
followed by a distillation under reduced pressure to a 330 mL
(1.65 mL/g) volume, were performed to remove the excess
diisopropylamine and methanol. The mixture was diluted with
tert-amyl alcohol (1270 mL, 6.35 mL/g) to give a tert-amyl
alcohol solution of the product 2-[3-(diisopropylamino)-1-
phenylpropyl]-4-(hydroxymethyl)phenol rac-2 for use in the
next step. Quantitative HPLC analysis indicated the crude
solution contained 221 g of product (83% yield). This inter-
mediate is not isolated and carried through to the next step with
no further purification.
Figure 5. Potential acylation products.
synthesis of fesoterodine. Because this route does not require the
use of protecting groups, redox chemistry. or functional group
manipulations, it is significantly more efficient than the existing
process (6 vs 11 steps and only 4 isolations), thus allowing
significant environmental savings (green chemistry E factor (298
vs 711) and solvent usage (280 kg/kg vs 644 kg/kg) improved by
60%) .
’ EXPERIMENTAL SECTION
General. Melting points were determined by closed cell DSC.
All reagents and solvents were used as received without further
purification.
(R)-2-[3-(diisopropylamino)-1-phenylpropyl]-4-(hydroxy-
methyl)phenol (R)-acetoxy(phenyl)acetate (18). A solution of
2-[3-(diisopropylamino)-1-phenylpropyl]-4-(hydroxymethyl)-
phenol in tert-amyl alcohol (44.2 L, equivalent to 2.95 kg of rac-2,
8.64 mol, 1 equiv) was heated to 70 °C. (R)-(À)-O-Acetylman-
delic acid (0.84 kg, 4.32 mol, 0.5 equiv) was dissolved in tert-amyl
alcohol (14.8 L), and the resulting solution was added to the
solution of rac-2 in tert-amyl alcohol keeping the internal
temperature at 70 °C. The solution was seeded with 18 (0.03 kg,
1 wt %). The resulting slurry was cooled to 60 °C over 2 h and
then to 25 °C over another 3 h. The mixture was stirred at 25 °C
for an additional 12 h. The slurry was filtered, and the cake was
deliquored well. The cake was slurry washed with tert-amyl
alcohol (2 Â 29.5 L, 2 Â 10 mL/g) and deliquored well. The
white solid was dried under reduced pressure at 40 °C for 12 h.
(R)-2-[3-(diisopropylamino)-1-phenylpropyl]-4-(hydroxymethyl)-
phenol (R)-acetoxy(phenyl)acetate 18 (2.04 kg, 3.81 mol) was
isolated in 37.8% yield (corrected for 14.3% w/w tert-amyl
(2-Hydroxy-4-phenyl-3,4-dihydro-2H-chromen-6-yl)meth-
anol (4). 4-(Hydroxymethyl)phenol 5 (100 g, 0.81 mol, 1.0 equiv)
was stirred with N-methylpiperazine (202 g, 2.01 mol, 2.5 equiv)
in toluene (900 mL, 9 mL/g) and then heated to reflux. Upon
reaching reflux, cinnamaldehyde 6 (133 g, 1.01 mol, 1.25 equiv)
was then added over 3 h maintaining the reaction mixture at
reflux with azeotropic removal of water. Once the addition was
complete the reaction mixture was heated at reflux with removal
of water for 2 h, and then some toluene was removed by distilling
under reduced pressure, reducing the volume to approximately
600 mL. The mixture was then allowed to cool to room
temperature, and ethyl acetate (1.8 L, 18 mL/g) was added. The
product solution was then sequentially washed with 2 M aqueous
hydrochloric acid (1.8 L, 18 mL/g), 1 M aqueous hydrochloric acid
(700 mL, 7 mL/g), 0.25 M aqueous sodium hydrogen carbonate
solution (700 mL, 7 mL/g), and water (1 L, 10 mL/g). The
organic phase was then diluted with toluene (650 mL, 6.5 mL/g),
and the mixture was distilled down to approximately 600 mL
volume. The mixture was cooled to 22 °C and stirred for 6 h. The
suspension was cooled to 2 °C and stirred for a further 2 h. The
slurry was filtered, and the cake was washed twice with toluene
(300 mL, followed by 100 mL). The resulting pale tan solid was dried
in vacuum for 24 h at up to 60 °C, to give (2-hydroxy-4-phenyl-
3,4-dihydro-2H-chromen-6-yl)methanol 4 (118.6 g, 57% yield).
1H NMR (400 MHz, DMSO-d6) δ ppm 7.35 (m, 2 H); 7.23
(m, 4 H); 7.06 (m, 1 H); 6.78 (m, 1 H); 6.64 (bs, 0.76 H); 6.56
(bs, 0.24 H); 5.54 (m, 0.76 H); 5.44 (m, 0.24 H); 4.93 (m, 1 H);
4.28 (m, 3 H); 2.08 (m, 2 H). 13C NMR (400 MHz, DMSO-d6)
δ ppm Major diastereoisomer 151.3, 144.6, 134.0, 128.6, 128.5,
127.4, 126.5, 126.2, 124.9, 116.3, 90.4, 62.6, 36.7, 36.5. Minor
diastereoisomer 152.7, 144.3, 134.1, 128.6, 128.5, 127.1, 126.6,
126.2, 125.2, 116.1, 94.1, 62.6, 36.7, 36.5. HRMS (ES) Calcd for
C16H16O3Na (MNa+) 279.0992, found 279.0994.
1
alcohol) and 99% ee. H NMR (400 MHz, DMSO-d6) δ ppm
7.42À7.46 (m, 2 H); 7.23À7.35 (m, 7 H); 7.13À7.19 (m, 2 H);
6.95 (dd, J = 8.21, 2.15 Hz, 1 H); 6.76 (d, J = 8.21 Hz, 1 H); 5.64
(s, 1 H); 4.29À4.38 (m, 3 H); 3.32 (br. s., 2 H); 2.54À2.76 (m, 2
H); 2.29 (br. s., 2 H); 2.06 (s, 3 H); 1.37 (d, J = 7.62 Hz, 1 H);
1.01À1.09 (m, 13 H). 13C NMR (400 MHz, DMSO-d6) δ ppm
170.1, 169.6, 153.6, 144.3, 137.1, 132.8, 129.6, 128.1, 128.0,
127.8, 127.7, 127.5, 26.1, 125.8, 125.6, 114.8, 76.3, 68.8, 63.0,
44.6, 40.9, 35.9, 28.7, 20.8, 8.6. Diacel Chiralpak IC 250 mm Â
4.6mm, 5 μm; isochratic 96:4 heptane/ethanol + 0.5% diethyla-
mine, flow 1.0 mL/min, 210 nm, Rt (min) 12 (undesired), 13
(desired). Enantiomeric excess 99%. HRMS (ES) Calcd for
C22H32NO2 (MH+) 342.2428, found 342.2435.
(R)-2-[3-(Diisopropylamino)-1-phenylpropyl]-4-(hydroxy-
methyl)phenol (2). (R)-2-[3-(diisopropylamino)-1-phenyl-
propyl]-4-(hydroxymethyl)phenol (R)-acetoxy(phenyl)acetate
18 (30 g, 0.056 mol, 1.0 equiv) was slurried in toluene
(180 mL, 6 mL/g) and warmed to 50 °C. A 10 wt %/vol aqueous
2-[3-(Diisopropylamino)-1-phenylpropyl]-4-(hydroxy-
methyl)phenol (rac-2). (2-Hydroxy-4-phenyl-3,4-dihydro-
2H-chromen-6-yl)methanol 4 (200 g, 0.78 mol, 1.0 equiv) was
1015
dx.doi.org/10.1021/op200107g |Org. Process Res. Dev. 2011, 15, 1010–1017

Organic Process Research & Development
ARTICLE
solution of potassium carbonate (180 mL, 6 mL/g) was charged
maintaining the temperature at 50 °C. The mixture was stirred
vigorously at 50 °C for 6 h. The two solution phases were allowed
to settle and were separated at 50 °C. The organic phase was
washed with water (120 mL, 4 mL/g) at 50 °C. The phases were
separated at 50 °C, and the toluene volume was reduced to
120 mL (4 mL/g) by distillation under reduced pressure. The
temperature was adjusted to 60 °C and then cooled to 40 °C over
1 h. The batch was held at 40 °C and then seeded with 2 (150 mg).
The mixture was granulated for 90 min at 40 °C and then cooled
to 20 °C over 4 h. The batch was granulated at 20 °C for 4 h. The
slurry was then cooled to 2 °C over 2 h and granulated at 2 °C for
4 h. The suspension was filtered, the cake was washed with cold
toluene (30 mL, 1 mL/g), and the resulting white solid was dried
at 35 °C for 12 h to give (R)-2-[3-(diisopropylamino)-1-phenyl-
propyl]-4-(hydroxymethyl)phenol 2 (16.43 g) in 85.9% yield. 1H
NMR (400 MHz, DMSO-d6) δ ppm 9.19 (br. s., 1 H); 7.21À
7.32 (m, 4 H); 7.17 (d, J = 1.95 Hz, 1 H); 7.09À7.15 (m, 1 H);
6.94 (dd, J = 8.11, 2.05 Hz, 1 H); 6.72 (d, J = 8.21 Hz, 1 H); 4.93
(br. s., 1 H); 4.32À4.41 (m, 3 H); 2.95 (dt, J = 13.09, 6.55 Hz, 2 H);
2.27À2.37 (m, 2 H); 2.06 (q, J = 7.23 Hz, 2 H); 0.89 (dd, J = 6.45,
1.37 Hz, 12 H). 13C NMR (400 MHz, DMSO-d6) δ ppm 153.4,
145.3, 132.7, 130.8, 128.1, 127.9, 127.6, 126.1, 125.7, 125.5, 125.2,
114.6, 63.0, 47.8, 43.0, 40.6, 36.1, 20.7, 20.5. IR (KBr pellets) cmÀ1
3142, 3082, 3025, 2975, 2936, 2868, 1610, 1491, 1452, 1438, 1388,
1366, 1269, 1242, 1110, 1011, 904, 767, 744, 698. HRMS (ES)
Calcd for C22H32NO2 (MH+) 342.2428, found 342.2435.
(R)-(+)-Isobutyric Acid 2-[3-(Diisopropylamino)-1-phenyl-
propyl]-4-(hydroxymethyl)phenyl Ester 1. (R)-2-[3-(Diisopro-
pylamino)-1-phenylpropyl]-4-(hydroxymethyl)phenol 2 (50 g,
0.146 mol, 1.0 equiv) was dissolved in dichloromethane (400 mL,
8 mL/g) and then cooled to À12 °C. To this was added a
solution of isobutyryl chloride (16.15 g, 0.155 mol, 1.04 equiv) in
dichloromethane (250 mL, 5 mL/g), maintaining the reaction
temperature between À15 and À10 °C, followed by a vessel and
line rinse of dichloromethane (100 mL, 2 mL/g). The reaction
mixture was stirred at À12 °C for 2 h. A 5 wt %/wt aqueous
sodium carbonate solution (110 mL, 2.2 mL/g) was then added
to the reaction, allowing the temperature to rise towards 0 °C
during the addition, and the resulting pH was confirmed to be
between pH 7.5 and 8.5. The two phases were allowed to settle,
and the organic phase was sequentially washed with water
(450 mL, 9 mL/g), 5% wt/wt aqueous sodium carbonate
solution (450 mL, 9 mL/g), and twice with water (2 Â
450 mL, 2 Â 9 mL/g). The product solution was then concen-
trated under reduced pressure to a volume of 260 mL, and methyl
ethyl ketone (500 mL, 10 mL/g) was added. The solution was
reconcentrated under reduced pressure to a volume of 260 mL.
Two further additions of methyl ethyl ketone (each of 500 mL,
10 mL/g), each followed by a distillation under reduced pressure
to 260 mL, were performed to remove the dichloromethane. This
provided a methyl ethyl ketone solution of the product (R)-
(+)-isobutyric acid 2-[3-(diisopropylamino)-1-phenylpropyl]-
4-(hydroxymethyl)phenyl ester 1 for use in the next step
(fumarate salt formation). Quantitative HPLC analysis indicated
the solution contained 52.85 g of product (88% yield).
0.131 mol, 1.0 equiv) in methyl ethyl ketone (270 mL, 5 mL/g)
was then added to the slurry of fumaric acid followed by a vessel
and line rinse with methyl ethyl ketone (38 mL, 0.7 mL/g). The
resulting mixture was warmed to 37 °C with agitation for 30 min
ensuring that all the solids fully dissolved. The solution was
filtered into a crystallising vessel, rinsing the vessel and lines with
methyl ethyl ketone (70 mL, 1.3 mL/g). The solution was cooled
to 20 °C and seeded with 1-fumarate (0.54 g). After holding the
mixture at 20 °C for 1 h the slurry was cooled to 5 °C, and filtered
cyclohexane (65 mL, 1.2 mL/g) was added over 1 h. The mixture
was stirred at 5 °C for 8 h and then filtered, washing the cake with
a mixture of cyclohexane (65 mL) and methyl ethyl ketone
(16 mL), followed by cyclohexane (54 mL), and the product was
dried under vacuum at 22 °C for 16 h to give (R)-(+)-isobutyric
acid 2-[3-(diisopropylamino)-1-phenylpropyl]-4-(hydroxymethyl)-
phenyl ester hydrogen fumarate 1-fumarate (57.1 g, 82% yield).
1H NMR (400 MHz, DMSO-d6) δ ppm 7.71 (br. s., 1 H); 7.26
(m, 4 H); 7.17 (d, J = 8.3 Hz, 1 H); 7.15 (m, 1 H); 6.91 (d, J =
8.3 Hz, 1 H); 6.87 (s, 2 H); 4.84 (br. s., 1 H); 4.62 (d, J = 13.3
Hz, 2 H); 3.96 (t, J = 8.4 Hz, 1 H); 3.63 (sept, J = 6.6 Hz, 2 H);
2.96 (br. s., 1 H); 2.81 (sept, J = 7.0 Hz, 1 H); 2.73 (br. s., 1 H);
2.86À2.57 (m, 4 H); 1.33 (d, J = 7.0 Hz, 6 H); 1.27 (d, J = 6.6
Hz, 12 H). 13C NMR (400 MHz, DMSO-d6) δ ppm 175.6,
170.3, 147.5, 142.3, 140.5, 135.6, 134.4, 128.7, 127.6, 126.8,
126.7, 126.4, 122.2, 63.7, 54.4, 45.8, 41.9, 34.1, 31.6, 19.1, 18.9,
17.7. Anal. Calcd: C (68.29%), H (7.83%), N (2.65%). Found:
C (68.40%), H (7.92%), N (2.58%). IR (KBr pellets) cmÀ1
3473, 2978, 2937, 2877, 2825, 2769, 2696, 1755, 1707, 1570,
1468, 1387, 1234, 1178, 1097, 984, 912, 868, 795, 748, 702.
Mp 103.5 °C.
’ AUTHOR INFORMATION
Corresponding Author
*E-mail: Dirat_olivier@yahoo.co.uk.
’ ACKNOWLEDGMENT
The authors thank Simon Davies, Paul McDaid and Pat
O’Neill from the Pfizer Process Development Center in
Loughbeg (Ireland) for early development work on the
preparation of the lactol; Ciaran Byrne, Mairead Kelleher,
Dennis Lynch, Will Mullally, Sandra Mullane from the Pfizer
API manufacturing plant in Ringaskiddy (Ireland) for running
this process at commercial scale; Robert Bright and Andrew
Fowler from the Sandwich Pilot Plant for scaling-up this
process; Sophie Apoux, Francois Lestremeau and Maura
Wood for help with elucidation of structures; John Deering
and Trevor Newbury for technical assistance with hydrogena-
tion reactions; Charles Gordon and Wilfried Hoffmann for
engineering assessment; Neal Sach and Katie Wilford for
reaction and solubility screening; Christian Regius for process
safety evaluation; Lynsey Hesmondhalgh and Rosalind Sankey
for work on the final steps of the process; Kathryn Arthur,
Steve Belsey, Samantha Johnson, Tony Stevens and Sophie
Vernay for analytical support and Julian Smith for reviewing
and improving this manuscript.
(R)-(+)-Isobutyric Acid 2-[3-(Diisopropylamino)-1-phenyl-
propyl]-4-(hydroxymethyl)phenyl Ester Hydrogen Fuma-
rate 1-Fumarate. Fumaric acid (14.47 g, 0.125 mol, 0.95 equiv)
was slurried in methyl ethyl ketone (162 mL, 3 mL/g) at 20 °C. A
solution of (R)-(+)-isobutyric acid 2-[3-(diisopropylamino)-
1-phenylpropyl]-4-(hydroxymethyl)phenyl ester 1 (54.03 g,
’ REFERENCES
(1) McKeage, K.; Keating, G. M. Drugs 2009, 69 (6), 731.
(2) Malhotra, B.; Gandelman, K.; Sachse, R.; Wood, N.; Michel,
M. C. Curr. Med. Chem. 2009, 16, 4481.
1016
dx.doi.org/10.1021/op200107g |Org. Process Res. Dev. 2011, 15, 1010–1017

Organic Process Research & Development
ARTICLE
(3) (a) Browne, R.; Kilkelly, M. Patent WO 2007/137799 A1, 2007.
(b) Meese, C. Patent WO 2007/144097 A1, 2007.
(4) The formation of quinone methide 20 is achieved either in acidic
media (Brønsted or Lewis) or by heat.
(5) (a) Paras, N. A.; MacMillan, D. W. C. J. Am. Chem. Soc. 2001,
123, 4370. (b) Bartoli, G.; Melchiorre, P. Synlett 2008, 1759. (c) Paras,
N. A.; Simmons, B.; MacMillan, D. W. C. Tetrahedron 2009, 65, 3232.
(6) List, B. Angew. Chem., Int. Ed. 2010, 49, 1730.
(7) (a) Mills, I. W.; Scholfield, D. P. Patent WO 2010/018484 A1,
2010. (b) Ahman, J. B.; Dillon, B. R.; Pettman, A. J. Patent WO 2007/
138440 A1, 2007.
(8) For an example of a LangmuirÀHinshelwood kinetic model
applied to heterogeneous hydrogenation, see: Vincent, M. J.; Gonzalez,
R. D. Appl. Catal., A 2001, 217, 143.
(9)
1017
dx.doi.org/10.1021/op200107g |Org. Process Res. Dev. 2011, 15, 1010–1017